Family Medicine Physicians

29-1215.00
Median wage $244,180/yr107,510 employed (US)Rank #665 of 923 scored · top 72% by substitution

Diagnose, treat, and provide preventive care to individuals and families across the lifespan. May refer patients to specialists when needed for further diagnosis or treatment.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution20
Exposure20
Augmentation65

Substitution — the headline: capability discounted by cost, barriers and adoption.

Exposure — technical capability alone, regardless of whether anyone deploys it.

Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.

Tasks on the substitution scale

12 rated tasks, binned by substitution score.

Position among all scored occupations

Distribution of 923 occupation scores; the marker is this occupation.

Tasks with substitution ≥ 70

0%

Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.

Why this score

The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.

Task automatabilityw 35%19

panel mean rating 1.7/5 → substitution pressure 19/100

Technical feasibility todayw 20%23

panel mean rating 1.9/5 → substitution pressure 23/100

Cost vs. human wagew 15%24

panel mean rating 2.0/5 → substitution pressure 24/100

Adoption barriersw 20%inverted — strong barriers lower the score14

panel mean rating 4.4/5 (barrier strength) → substitution pressure 14/100

Sector adoption velocityw 10%26

panel mean rating 2.0/5 → substitution pressure 26/100

Task breakdown (12 tasks)

Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.

Collect, record, and maintain patient information, such as medical history, reports, or examination results.

49

CI 36–61 · exposure 47 · augmentation 88 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Healthcare is digitizing rapidly and AI-based documentation and data-capture tools are in active deployment across primary care and hospital systems. Adoption is driven by administrative burden reduction and EHR vendor integration, though uptake varies by practice size and setting.
Sector adoption velocityclaude-sonnet-53/5Healthcare has lagged other information sectors in AI adoption due to regulatory and liability concerns, but ambient documentation tools have seen rapid uptake in the last two years among large health systems, putting it in a middling-to-accelerating category.
Augmentation potentialclaude-haiku-4-5-202510015/5AI-powered transcription, auto-population of forms, and intelligent summarization of past records substantially improve a physician's productivity in documentation and record review while the physician remains responsible for accuracy and clinical decisions. This is one of the most mature augmentation use cases in clinical practice today.
Augmentation potentialclaude-sonnet-54/5AI ambient scribes and transcription tools substantially reduce documentation burden and time spent on notes, letting physicians focus more on patient interaction while AI drafts records for review.
Task automatabilityclaude-haiku-4-5-202510013/5AI can automate significant portions of data entry and structuring from dictations, images, and lab results into electronic health records, but requires human review for accuracy and completeness. Achieving 50% time savings on the full task (collecting, recording, and maintaining across diverse patient sources) is feasible with AI-assisted EHR systems, though setup and oversight are nontrivial.
Task automatabilityclaude-sonnet-52/5AI can assist with ambient scribing and transcription of history-taking, but collecting information via physical exam, patient rapport-building, and clinical judgment about what to record still require human involvement, so full end-to-end automation is not yet achieved.
Adoption barriersclaude-haiku-4-5-202510013/5While regulatory requirements (HIPAA, medical documentation standards) and organizational friction exist, they do not legally mandate human recording—many tasks can be delegated to administrative staff or AI. However, physician accountability for chart accuracy and potential liability for AI errors create material adoption friction.
Adoption barriersclaude-sonnet-54/5Medical record-keeping is heavily regulated (HIPAA, medical licensing, liability for errors in clinical documentation), and only a licensed clinician can verify and be accountable for the accuracy of patient records.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI-driven EHR and transcription services cost substantially less than physician time for data entry and administrative work. The marginal cost of processing patient information through AI is typically orders of magnitude cheaper than paying a physician or scribe to do the same work manually.
Cost vs. human wageclaude-sonnet-53/5AI scribe/documentation tools cost a subscription fee per provider that is much less than a scribe's salary, but physician time for review and correction plus integration with EHR systems keeps net savings moderate rather than order-of-magnitude cheaper for the whole task.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed EHR systems with AI-powered transcription, documentation templates, and structured data capture are in widespread production use in clinics and hospitals. However, error rates in transcription and data interpretation remain material enough that human verification is standard practice, preventing a perfect 5 rating.
Technical feasibility todayclaude-sonnet-53/5Ambient AI scribes (e.g., Nuance DAX, Abridge) are deployed in production at many health systems and reliably generate structured notes from visits, but they still require physician review and correction, and don't perform the actual data collection (exam, history elicitation) autonomously.

Prepare government or organizational reports which include birth, death, and disease statistics, workforce evaluations, or medical status of individuals.

31

CI 25–37 · exposure 33 · augmentation 63 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Health systems are slowly adopting EHR-integrated reporting tools, but most physician practices still rely on manual report preparation or generic data exports; widespread production adoption of AI-driven clinical report generation remains limited due to regulatory caution and liability concerns.
Sector adoption velocityclaude-sonnet-52/5Healthcare administrative documentation is adopting AI tools at a moderate pace, but formal statistical/government reporting workflows remain slow to change due to regulatory and liability constraints.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by auto-populating data fields, flagging anomalies in statistics, and generating report templates, allowing physicians to focus on interpretation and validation rather than data entry—useful productivity gain within a supervised workflow.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully speed up data aggregation, drafting, and formatting of these reports, letting physicians focus on verification and judgment calls rather than manual compilation.
Task automatabilityclaude-haiku-4-5-202510012/5While data aggregation and statistical compilation from structured records can be partially automated, family medicine physicians must interpret clinical context, ensure accuracy of sensitive health data, and make judgment calls about what statistics to highlight—tasks requiring human oversight that prevent the ≥50% time-saving threshold from being consistently met.
Task automatabilityclaude-sonnet-53/5AI can draft report narratives and compile statistics from structured data, but pulling accurate data from varied sources, verifying clinical accuracy, and finalizing for regulatory submission still requires substantial human involvement.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: physicians must often legally sign off on vital statistics and disease reports; regulatory requirements (CDC, state health departments) mandate authorized personnel; HIPAA and other privacy regulations create liability asymmetry if automation produces errors, making human accountability non-substitutable.
Adoption barriersclaude-sonnet-54/5These reports often require licensed physician certification (e.g., death certificates, workforce/medical status attestations) with legal accountability, making full automation and sign-off substitution difficult.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for data extraction and basic report templating exist but require significant setup, data cleaning, and physician review/validation; the total cost including these integration steps and the mandatory physician oversight approaches the loaded cost of the physician performing the task partially themselves.
Cost vs. human wageclaude-sonnet-53/5AI can cut drafting time significantly, but physician verification, data integration with EHR/registry systems, and compliance checks keep overall cost savings moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510012/5Current AI systems can extract and organize structured data from medical records, but deployed products rarely handle the full complexity of multi-source health data integration, regulatory compliance (HIPAA, vital statistics reporting), and the clinical interpretation required for authoritative government/organizational reports.
Technical feasibility todayclaude-sonnet-52/5AI drafting tools and EHR-integrated report generators exist but are not widely deployed for producing final government/organizational health statistics reports without heavy physician review and correction.

Advise patients and community members concerning diet, activity, hygiene, and disease prevention.

27

CI 25–29 · exposure 25 · augmentation 63 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for autonomous patient counseling remains limited; most healthcare organizations use AI as a supplementary tool (e.g., patient education content) rather than as a replacement, reflecting both regulatory caution and clinical resistance to removing physician-patient interaction.
Sector adoption velocityclaude-sonnet-52/5Healthcare is a relatively slow-adopting sector for patient-facing AI advice due to regulatory caution, liability concerns, and the sensitivity of clinical guidance, with pilots more common than production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by generating evidence-based dietary and lifestyle suggestions, summarizing guidelines, and creating customized educational materials that physicians review and tailor, meaningfully reducing preparation time while the physician retains clinical authority and patient relationship.
Augmentation potentialclaude-sonnet-54/5AI tools can draft patient education materials, summarize guidelines, and suggest talking points, meaningfully speeding up how physicians prepare and deliver counseling while they retain responsibility for accuracy and delivery.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate standardized dietary and hygiene guidance, the task requires personalized medical advice tailored to individual patient histories, comorbidities, and preferences—elements that demand human judgment and cannot be fully automated to meet the ≥50% time-saving threshold without physician oversight.
Task automatabilityclaude-sonnet-52/5AI chatbots can generate generic diet, hygiene, and prevention advice, but personalized counseling that integrates a patient's history, comorbidities, and rapport-building requires physician judgment and trust that current systems cannot fully replace end-to-end.the task also includes relational and motivational components not captured by text generation alone.
Adoption barriersclaude-haiku-4-5-202510014/5Strong legal and liability barriers protect this task: physicians bear professional and legal responsibility for patient advice; medical malpractice liability, regulatory oversight, and the expectation of a licensed physician–patient relationship create hard constraints on full substitution.
Adoption barriersclaude-sonnet-54/5Medical advice-giving is closely tied to licensed practice, malpractice liability, and standard-of-care requirements, meaning most jurisdictions and health systems require a licensed provider to be involved in advice given to patients.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-assisted chatbots or educational modules have low per-interaction costs, but they do not eliminate the physician's need to review, customize, and validate advice for each patient, making the all-in cost per medically safe counseling encounter remain substantial relative to the human alternative.
Cost vs. human wageclaude-sonnet-53/5Generating generic advice text is cheap via AI, but the full task requires clinical contextualization and physician verification, keeping oversight costs comparable to or only modestly below the cost of physician time.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs end-to-end patient counseling on diet and disease prevention autonomously; products exist for patient education and reminders but require significant physician supervision and cannot replace the clinical assessment and individualization physicians provide.
Technical feasibility todayclaude-sonnet-52/5Consumer health chatbots and patient portals offer general wellness advice today, but no deployed product independently delivers physician-level personalized preventive counseling at scale without human oversight due to liability and accuracy concerns.

Refer patients to medical specialists or other practitioners when necessary.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While health systems are piloting decision-support tools, adoption of AI-driven referral systems remains limited and cautious. Most adoption remains in pilot or experimental phases rather than deep production deployment across the sector.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall is a slower-adopting sector for autonomous clinical decision-making due to regulatory, liability, and EHR integration hurdles, though administrative AI adoption is growing.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can helpfully surface referral candidates, flag guideline-based criteria, and reduce cognitive load in identifying when specialty input is needed, but the physician remains the final arbiter of appropriateness and execution.
Augmentation potentialclaude-sonnet-54/5AI can help physicians by summarizing patient records, suggesting relevant specialists, flagging red-flag symptoms warranting referral, and drafting referral documentation, meaningfully speeding the process while the physician retains decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI systems can assist in identifying when specialist referral is appropriate by analyzing clinical data and guidelines, the final referral decision requires clinical judgment, patient context, and practitioner accountability. Current AI cannot reliably handle the full spectrum of clinical complexity and liability considerations that mandate human physician oversight.
Task automatabilityclaude-sonnet-52/5Referral decisions require clinical judgment integrating patient history, exam findings, and risk assessment; AI can assist with drafting referral letters or suggesting specialists but cannot independently decide and execute referrals reliably today.
Adoption barriersclaude-haiku-4-5-202510014/5Medical licensure, malpractice liability, and regulatory requirements (e.g., standard of care) mandate that a licensed physician make or authorize referral decisions. Legal and professional standards effectively prevent autonomous AI substitution.
Adoption barriersclaude-sonnet-55/5Only a licensed physician can legally make referral decisions and bear liability for missed or inappropriate specialist referrals, making this a hard regulatory and licensure barrier.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-assisted referral systems have implementation and oversight costs, but the core task (clinical judgment and decision-making) remains physician-centric. The loaded cost of physician review and validation remains comparable to or exceeds the marginal benefit of AI tools.
Cost vs. human wageclaude-sonnet-52/5Physician time for referral decisions is a small part of a visit, and AI tools still require physician review and sign-off, so cost savings are modest relative to liability and integration overhead.
Technical feasibility todayclaude-haiku-4-5-202510012/5Decision-support tools exist to flag potential specialist referrals, but no deployed product independently makes and executes referral decisions in clinical production at scale. These systems require substantial human review and final authorization by the physician.
Technical feasibility todayclaude-sonnet-52/5Some clinical decision support tools flag referral needs or suggest specialists, but no deployed product autonomously manages the referral decision and process at scale in production.

Order, perform, and interpret tests and analyze records, reports, and examination information to diagnose patients' condition.

21

CI 20–23 · exposure 25 · augmentation 75 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Healthcare organizations are actively piloting AI diagnostic support tools and clinical decision systems, but adoption remains cautious and limited to structured subdomains (radiology, pathology). Broad production deployment of autonomous diagnostic AI remains uncommon due to liability and regulatory friction.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall adopts AI diagnostic tools cautiously due to regulatory, liability, and workflow integration barriers, with pilots more common than widespread production deployment for core diagnosis.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments diagnostic capability through rapid literature synthesis, pattern recognition across large datasets, differential diagnosis prompting, and image/lab interpretation support. Physicians using these tools can improve diagnostic accuracy and speed while maintaining clinical oversight and accountability.
Augmentation potentialclaude-sonnet-54/5AI meaningfully assists physicians via differential diagnosis generators, lab/imaging analysis, and clinical decision support, improving speed and thoroughness while the physician retains final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with interpreting imaging and some lab tests, the task requires integrated clinical judgment across multiple data streams, patient history synthesis, and accountability for diagnostic accuracy. Current systems excel at narrow subtasks (e.g., radiology interpretation) but cannot reliably execute the full end-to-end diagnostic workflow with required quality and safety standards.
Task automatabilityclaude-sonnet-52/5AI can assist with test interpretation and differential diagnosis suggestions, but the full end-to-end task—ordering appropriate tests, physical examination integration, and final diagnostic judgment across undifferentiated patients—requires clinical reasoning and physical assessment AI cannot yet perform independently at equal quality.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and regulatory frameworks require a licensed physician to take responsibility for diagnosis and clinical decision-making; this is a foundational standard in medical practice law across jurisdictions. Malpractice liability, patient safety obligations, and licensure requirements create hard barriers to autonomous automation.
Adoption barriersclaude-sonnet-55/5Diagnosis is a core licensed medical act with direct malpractice liability; regulations require a physician to make and be accountable for diagnostic determinations.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integration of AI diagnostic tools into clinical workflows carries significant setup, validation, and oversight costs. Per-patient inference costs remain low, but total implementation and liability costs are substantial relative to the physician's loaded wage for this task.
Cost vs. human wageclaude-sonnet-52/5AI tools for narrow interpretation tasks (e.g., some imaging analysis) are cheap, but the physician's synthesis, exam, and liability-bearing diagnosis still require a costly licensed professional, keeping overall cost comparable to human-only workflows.
Technical feasibility todayclaude-haiku-4-5-202510012/5Diagnostic support tools exist (clinical decision support, image analysis) but typically operate in narrow domains and require physician validation. No deployed system reliably performs independent diagnosis at the quality threshold required for patient care; all production systems are assistive rather than autonomous.
Technical feasibility todayclaude-sonnet-52/5Deployed diagnostic decision-support tools and AI-assisted imaging/lab interpretation exist, but no production system independently orders and interprets the full diagnostic workup and delivers diagnoses reliably across primary care's broad case mix.

Explain procedures and discuss test results or prescribed treatments with patients.

21

CI 18–25 · exposure 25 · augmentation 63 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Despite digitization of healthcare, actual displacement of physician-patient discussion by AI in production remains minimal; adoption is limited to pilot projects and narrow use cases, not systemic replacement, reflecting both regulatory constraints and sector-specific resistance to automating human contact.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall adopts AI slowly for patient-facing clinical communication due to regulatory, liability, and trust concerns, though administrative and documentation uses are growing faster.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by pre-drafting explanatory materials, summarizing test results, or generating talking points, which can accelerate the physician's preparation for a discussion. However, the core interactive and interpersonal work remains human-centered, and augmentation is partial rather than transformative.
Augmentation potentialclaude-sonnet-54/5AI tools (e.g., ambient scribes, patient-friendly result summaries, decision support) increasingly help physicians prepare and personalize these conversations, meaningfully boosting efficiency while the physician remains the one delivering the explanation.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate explanatory text about procedures and treatments, this task requires real-time, individualized patient communication adapted to literacy level, emotional state, and specific medical history—capabilities that current AI systems cannot perform end-to-end reliably. The interactive, nuanced nature of patient discussion and the need for compassionate, contextual judgment are beyond the 50% time-saving threshold.
Task automatabilityclaude-sonnet-52/5AI can draft explanations of procedures or results, but the live, adaptive dialogue with a patient—reading emotional cues, answering follow-up questions, and tailoring communication—still requires a human physician for most of the interaction.
Adoption barriersclaude-haiku-4-5-202510014/5High barriers exist: patients have strong preference for physician-delivered explanations; liability and malpractice exposure are severe if AI-generated information is incorrect or misunderstood; regulatory and standard-of-care frameworks require physician presence and accountability; informed consent is a legal and ethical requirement tied to the physician.
Adoption barriersclaude-sonnet-55/5Discussing diagnoses, test results, and treatment plans is a core licensed medical activity with informed-consent and liability requirements, legally requiring a physician or supervised clinician.
Cost vs. human wageclaude-haiku-4-5-202510012/5An AI system capable of reliably handling patient consultation with proper compliance, oversight, and integration would require substantial infrastructure and human supervision, making total cost-per-task comparable to or exceeding the physician's marginal time on routine explanation—especially when factoring in liability and quality assurance.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply generate draft explanations, but the physician's time explaining and answering questions in person remains billed and required, so overall cost savings are modest given liability and oversight needs.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs full patient consultation and explanation at the standard of care required in medicine. Chatbots exist but lack the clinical judgment, legal accountability, and patient rapport necessary for actual clinical use; they remain experimental and not production-standard in real practices.
Technical feasibility todayclaude-sonnet-52/5Some patient portals use AI-generated summaries of lab results or chatbots for basic Q&A, but no deployed product reliably conducts the full clinical conversation about treatment or test results in place of a physician.

Train residents, medical students, and other health care professionals.

21

CI 16–25 · exposure 17 · augmentation 63 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Healthcare and medical education have traditionally lagged in AI adoption of tasks requiring direct human responsibility and judgment. While some institutions pilot AI-assisted learning tools, production-scale displacement of physician trainers remains limited and slow.
Sector adoption velocityclaude-sonnet-52/5Healthcare and medical education are historically slow to adopt automation for core supervisory/teaching functions, though AI-assisted study tools are gaining some traction.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist trainers by generating case scenarios, automating progress tracking, and providing evidence-based resources, moderately reducing preparation time. However, the core work of mentoring, real-time correction, and clinical role modeling remains largely resistant to meaningful augmentation.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully augment teaching by generating case studies, quizzes, differential diagnosis exercises, and personalized learning materials, enhancing physician-led instruction significantly.
Task automatabilityclaude-haiku-4-5-202510011/5Training medical professionals requires real-time feedback, socratic questioning, adaptive instruction based on individual learner understanding, and modeling of clinical judgment—capabilities current AI systems cannot replicate end-to-end. While AI can generate lectures or practice questions, the core pedagogical work of training remains fundamentally human.
Task automatabilityclaude-sonnet-52/5Clinical teaching involves live patient interactions, bedside mentoring, and personalized feedback based on trainee performance, which current AI cannot conduct end-to-end; AI can supply supplementary materials but not replace the supervisory teaching relationship.
Adoption barriersclaude-haiku-4-5-202510014/5Medical education and credentialing are heavily regulated; accreditation bodies (ACGME, GME programs) mandate that training be provided by licensed physicians meeting specific experience requirements. Liability and professional standards create hard barriers to full automation.
Adoption barriersclaude-sonnet-54/5Medical education and supervision of trainees is governed by accreditation bodies (ACGME) and licensing requirements mandating physician oversight, creating strong institutional and regulatory barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI can reduce costs of some preparatory content creation, but supervision and mentoring by an experienced physician command significant market value (often $150–300+/hour loaded). AI cannot yet fully substitute for this supervision cost-effectively, as quality training requires expert judgment.
Cost vs. human wageclaude-sonnet-52/5AI content generation is cheap, but the substantive teaching value comes from licensed physician oversight and hands-on supervision, which AI cannot substitute, so cost savings are limited to peripheral materials.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed AI products can generate educational content and practice materials, but no production system reliably replaces the mentor role, clinical supervision, or adaptive feedback that constitutes meaningful training. AI tutoring systems exist but are narrow in scope and supplementary rather than autonomous.
Technical feasibility todayclaude-sonnet-52/5Some AI tutoring and case-simulation products exist for medical education, but no deployed system reliably replaces attending-led clinical teaching, rounds supervision, or competency assessment of trainees.

Monitor patients' conditions and progress and reevaluate treatments as necessary.

18

CI 11–25 · exposure 17 · augmentation 75 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite healthcare's high digitization, adoption of AI for autonomous monitoring/treatment decisions remains slow due to liability concerns, regulatory scrutiny, and physician resistance rooted in patient safety and professional accountability. Most deployed tools are narrowly scoped alerts and documentation aids, not agents making clinical adjustments independently.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall has been a slower adopter of AI for core clinical decision-making due to regulatory, liability, and workflow integration challenges, despite some pilots in remote monitoring and CDS.
Augmentation potentialclaude-haiku-4-5-202510014/5Current AI tools meaningfully augment physician productivity on this task by automating data aggregation, flagging abnormal results, generating clinical summaries, and suggesting evidence-based protocols for review—substantially reducing the cognitive load of comprehensive monitoring while the physician retains decision authority and final judgment.
Augmentation potentialclaude-sonnet-54/5AI-powered monitoring dashboards, predictive alerts, and summarization tools meaningfully help physicians track patient status and flag changes needing reevaluation, improving efficiency while the physician remains decision-maker.
Task automatabilityclaude-haiku-4-5-202510012/5Monitoring and initial assessment of routine vital signs and lab trends can be partially automated (e.g., alerts for abnormal values), but the core clinical judgment required to synthesize multi-dimensional patient data and adjust treatment plans requires nuanced human expertise that current AI cannot reliably replicate end-to-end. The task involves complex patient-specific decision-making that falls well short of the 50% time-saving threshold for full automation.
Task automatabilityclaude-sonnet-51/5This task requires ongoing clinical judgment, physical examination, and synthesis of nuanced patient context that cannot be fully delegated to AI systems today; only small sub-components like flagging abnormal lab trends are automatable.
Adoption barriersclaude-haiku-4-5-202510014/5Strong legal and regulatory barriers exist: physicians have statutory responsibility for patient care decisions, malpractice liability attaches to treatment choices, and standard-of-care requirements mandate physician judgment in monitoring and treatment adjustment. Many patients expect human continuity of care, and medical licensing law effectively reserves the core decision-making to licensed clinicians.
Adoption barriersclaude-sonnet-55/5Reevaluating and modifying treatment is a licensed medical act requiring physician sign-off, with strong liability, regulatory, and scope-of-practice barriers preventing full AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510012/5The specialized expertise required (domain knowledge, liability insurance, regulatory compliance) means that even partial automation would require significant integration costs and physician oversight, making the all-in cost comparable to or exceeding physician time on this task. AI reduces some clerical burden but does not approach order-of-magnitude cost advantage.
Cost vs. human wageclaude-sonnet-52/5While monitoring alerts and data aggregation tools are cheap to run, the physician's judgment and liability for treatment changes remain required, so overall cost savings versus the human physician are modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5While clinical decision support systems and EHR alerts exist in production, they function as narrow assistive tools rather than end-to-end performers of the full monitoring and reevaluation task. Autonomous AI systems making treatment adjustments without human oversight do not have demonstrated reliability in production medical settings, and liability/regulatory constraints mean this remains largely in the physician's domain.
Technical feasibility todayclaude-sonnet-52/5Clinical decision support tools and remote monitoring alerts exist in production (e.g., EHR-based flagging systems), but no deployed product independently monitors and reevaluates treatment plans without physician oversight.

Coordinate work with nurses, social workers, rehabilitation therapists, pharmacists, psychologists, and other health care providers.

16

CI 6–25 · exposure 13 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Healthcare remains slow in AI adoption for high-stakes coordination; pilot systems for communication/scheduling exist but deep production deployment of autonomous coordination across clinical teams is rare.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall adopts digital tools slowly due to regulation, EHR fragmentation, and liability concerns, though administrative AI tools are gradually appearing.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems demonstrably assist physicians by automating scheduling, generating summaries of team input, flagging missing consultations, and managing shared notes—substantially raising coordination productivity when the physician retains final decision authority.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by drafting referral notes, summarizing patient records for other providers, and streamlining communication logistics, improving physician efficiency in coordination tasks.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with scheduling, summarizing communications, and flagging coordination gaps, this task fundamentally requires negotiating priorities, resolving conflicts, and building trust across diverse professionals—judgment-heavy work that current systems cannot handle end-to-end at ≥50% time savings and equal quality.
Task automatabilityclaude-sonnet-51/5Care coordination requires real-time interpersonal negotiation, judgment calls across disciplines, and situational adaptation that current AI cannot perform end-to-end without a human physician directing it.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and professional standards require a licensed physician to coordinate and take responsibility for the care plan; liability for coordination failures rests with the physician, and team members expect accountable human leadership, creating hard barriers to unsupervised automation.
Adoption barriersclaude-sonnet-54/5Licensure, liability for care decisions, and the requirement that a physician direct patient care create strong barriers to full automation of this coordination role.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI for communication routing, summarization, or light scheduling is cheap, but the physician must still oversee and validate all coordination, so the per-task cost savings are minimal relative to the physician's fully-loaded wage.
Cost vs. human wageclaude-sonnet-52/5AI communication tools are cheap, but they can't replace the physician's coordinating role, so any cost savings are marginal relative to the human labor still required.
Technical feasibility todayclaude-haiku-4-5-202510011/5No production system reliably performs independent coordination of multidisciplinary care teams; AI tools exist for scheduling and information routing but cannot autonomously manage the interpersonal and clinical negotiation inherent in this task.
Technical feasibility todayclaude-sonnet-52/5Some products assist with scheduling, messaging, or care-team documentation, but no deployed system autonomously coordinates multidisciplinary clinical care in production.

Plan, implement, or administer health programs or standards in hospitals, businesses, or communities for prevention or treatment of injury or illness.

12

CI 3–21 · exposure 13 · augmentation 50 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for health program planning remains slow and limited to pilot projects in large academic medical centers; most hospitals and community health organizations still rely on physician-led committees and manual program development with minimal AI integration.
Sector adoption velocityclaude-sonnet-52/5While healthcare uses AI for specific analytics and decision support, administrative and programmatic leadership functions show minimal AI-driven displacement to date.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist physicians by automating literature reviews, population health analytics, and draft program templates, improving the quality and speed of evidence synthesis. However, the human physician must remain central to strategy, implementation, and stakeholder engagement.
Augmentation potentialclaude-sonnet-53/5AI can assist with data analysis, epidemiological modeling, and drafting policy documents to inform program design, offering moderate productivity gains while humans retain decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with data analysis, guideline synthesis, and program design recommendations, the task fundamentally requires medical judgment, stakeholder coordination, organizational change management, and accountability that physicians must provide. Current systems cannot autonomously plan, implement, and administer comprehensive health programs at 50% time savings with equal quality.
Task automatabilityclaude-sonnet-51/5This task requires strategic planning, stakeholder coordination, and organizational judgment across complex systems that current AI cannot execute end-to-end without extensive human direction.
Adoption barriersclaude-haiku-4-5-202510014/5Substantial regulatory, liability, and organizational barriers exist: physicians must legally authorize health programs, maintain accountability for clinical outcomes, and secure institutional buy-in. Hospitals and healthcare organizations require licensed physician involvement in policy and program administration.
Adoption barriersclaude-sonnet-55/5Program administration in healthcare settings typically requires licensed physician oversight, regulatory compliance, and legal accountability that cannot be delegated to AI systems.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI-driven program support tools are not yet cost-effective substitutes for physician expertise; the overhead of AI systems, integration, human oversight, and clinical validation typically exceeds the cost of physician time for this complex strategic task.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the human role here, so there is no meaningful cost comparison—human physicians remain necessary for program design and oversight.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs end-to-end health program planning and administration; AI tools exist for epidemiological analysis and guideline retrieval but cannot autonomously design, implement, or manage programs across diverse hospital, business, or community settings.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product plans or administers health programs autonomously; this remains a human administrative and clinical leadership function.

Prescribe or administer treatment, therapy, medication, vaccination, and other specialized medical care to treat or prevent illness, disease, or injury.

4

CI 0–7 · exposure 5 · augmentation 63 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Adoption of AI for autonomous prescribing and treatment administration is negligible. Healthcare remains cautious and heavily regulated; physicians retain gatekeeping authority, and liability concerns slow any shift toward automation.
Sector adoption velocityclaude-sonnet-52/5Healthcare is adopting AI for diagnostics and documentation, but actual prescribing/treatment authority remains untouched by automation due to regulatory and safety constraints.
Augmentation potentialclaude-haiku-4-5-202510013/5AI augmentation tools—such as drug interaction checkers, evidence-based guideline summaries, and patient history aggregation—do assist physicians and raise productivity on parts of the task. However, the core decision and accountability remain human.
Augmentation potentialclaude-sonnet-54/5AI substantially assists physicians via decision-support, drug interaction checking, treatment guideline retrieval, and documentation, improving efficiency while the physician retains final authority.
Task automatabilityclaude-haiku-4-5-202510011/5Prescribing and administering medical treatment requires nuanced clinical judgment, examination of individual patient context, and accountability for adverse outcomes. Current AI cannot independently perform the full chain of diagnosis→treatment selection→monitoring without substantial human oversight, falling well short of the 50% time-saving threshold.
Task automatabilityclaude-sonnet-51/5Prescribing and administering treatment requires clinical judgment, physical examination, and legal authority that AI cannot independently exercise; no current system can perform this end-to-end with equal quality.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and regulatory requirements mandate that a licensed physician must prescribe and administer medical treatment. Liability, malpractice exposure, patient safety regulations, and state medical boards create hard barriers that prevent non-licensed entities from performing this task autonomously.
Adoption barriersclaude-sonnet-55/5Prescribing medication and administering treatment is tightly regulated and legally restricted to licensed physicians (or supervised practitioners), with significant liability exposure and mandatory human accountability.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of implementing, validating, and maintaining AI systems for prescribing, plus required physician oversight and liability coverage, exceeds the loaded cost of a physician making the decision themselves. AI does not yet offer cost parity for this high-liability task.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physician in this task, so there is no valid cost comparison—human performance remains mandatory, making AI substitution cost irrelevant or effectively infinite.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with drug interaction checking and evidence summarization, no deployed product autonomously prescribes or administers treatment at scale. Existing clinical decision support tools operate in narrow domains and require physician validation; they are not end-to-end systems.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously prescribes or administers medical treatment; clinical decision-support tools exist only as aids to licensed physicians.

Direct and coordinate activities of nurses, students, assistants, specialists, therapists, and other medical staff.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5There is no adoption of AI for autonomous staff direction in medicine. The task remains exclusively human-performed because it requires legal authority, accountability, and dynamic interpersonal leadership that licensing and liability frameworks preserve.
Sector adoption velocityclaude-sonnet-52/5Healthcare adoption of AI for clinical documentation and diagnostics is growing, but AI-driven staff coordination/management remains essentially unexplored in practice.
Augmentation potentialclaude-haiku-4-5-202510012/5Scheduling tools and communication platforms can modestly assist with logistics, but they do not meaningfully augment the physician's core work of directing priorities, resolving conflicts, or making real-time coordination decisions that require clinical judgment and accountability.
Augmentation potentialclaude-sonnet-53/5AI can assist with scheduling, task tracking, and communication support among care teams, improving efficiency, but does not replace the coordinating judgment of the physician.
Task automatabilityclaude-haiku-4-5-202510011/5Directing and coordinating medical staff requires real-time judgment, interpersonal negotiation, conflict resolution, and contextual decision-making that cannot be meaningfully automated by current AI systems. This task fundamentally depends on human authority, accountability, and dynamic team leadership.
Task automatabilityclaude-sonnet-51/5Directing and coordinating multidisciplinary clinical staff requires real-time judgment, authority, and interpersonal leadership that current AI cannot perform end-to-end.'
Adoption barriersclaude-haiku-4-5-202510015/5Legal, regulatory, and organizational structures require a licensed physician to hold accountability for team direction and clinical decisions. Medical governance, licensing boards, and liability frameworks mandate human professional oversight of coordinated care activities.
Adoption barriersclaude-sonnet-55/5Directing licensed medical staff and delegating clinical responsibility is legally tied to the physician's license and accountability, creating hard regulatory and liability barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI systems cannot perform this task, so direct cost comparison is not applicable. The cost of any AI that might assist is negligible against the irreducible human labor of actual staff direction and accountability.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this managerial task, so no meaningful cost comparison favors AI; the physician's judgment and authority remain irreplaceable.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product performs end-to-end staff direction and coordination in clinical settings. While scheduling and communication tools exist, they do not replace the judgment-laden role of a physician directing diverse teams in real time.
Technical feasibility todayclaude-sonnet-51/5No deployed product manages or directs clinical teams; AI tools at best support scheduling or documentation, not leadership/coordination of staff.

Related occupations — Healthcare Practitioners & Technical

How to read this

A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.

What would change this score

New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.